Gerd Ritter
Ludwig Institute for Cancer Research, New York, NY

Immunogenicity of Recombinant Antibodies: "In-Vivo Veritas"

Because of their potential to target and mediate selective destruction of cancer cells, monoclonal antibodies (mAbs) have generated much interest as diagnostic and therapeutic agents. For therapeutic application, the immunogenicity of the injected antibody is of paramount importance. For example, mouse-derived monoclonal antibodies are highly immunogenic in humans and generally can be given only for a limited period of time before onset of a human anti-mouse Ig response (HAMA), which rapidly clears the murine antibody from the blood and may trigger anaphylactic reactions. To attenuate or abolish immunogenicity and to allow long-term treatment of patients without the interference of an anti-immunoglobulin response, various molecular technologies have been employed that reduce the amount of foreign Ig residues in a therapeutic antibody, a factor hypothesized to inversely correlate with the immunogenicity of an antibody. Antibodies obtained by molecular engineering include human-mouse chimeras, humanized versions and fully human antibodies produced in human Ig transgenic animals and from human Ig gene libraries. However, the CDR regions of all these antibodies are unique and may contain inherently antigenic determinants that potentially can become immunogenic no matter how an antibody was generated.

To date, only limited information is available on the immunogenicity of humanized therapeutic antibodies in cancer patients. Among the reasons is the methodological challenge of measuring unambiguously and with high sensitivity human anti-human Ig antibody responses (HAHA) in patients. Recently we have developed a novel, highly sensitive biosensor assay (BIACORE), allowing to quantitate binding of patient sera to a humanized antibody without the use of secondary serological reagents (Cancer Res. 61:6851, 2001). Using this technique, we were able to carry out a detailed monitoring of HAHA reactivities in more than 40 patients with colon cancer enrolled in phase I and phase II clinical studies with the humanized antibody A33 (Ab A33 recognizes a 43k cell surface glycoprotein expressed in colon cancer; JCO 8:1894, 1990; PNAS 94:469, 1997). Patients in these studies were injected repeatedly with huAb A33 alone, with huAb A33 in combination with chemotherapy, or with radioisotope-labeled huAb A33 ([131I]-huAb A33) (details see Welt et al., abstract S01).

Before treatment with huAb A33, none of the patients with colon cancer had any measurable HAHA levels against huAb A33 or against other humanized control antibodies. After treatment with huAb A33, we found that 63 % (26/41) of the patients treated with repeated doses of huAb A33 developed HAHA against a conformational antigenic determinant located in the VL and VH regions of huAb A33. Detailed serological analysis showed two distinct types of HAHA. The two types differed with regard to the time point of HAHA onset, the course of HAHA after subsequent treatment with huAb A33, and in their impact on continued injection of huAb A33. A HAHA response of type I (49 % of patients) was characterized by an early onset with peak serum levels detectable after approximately two weeks of treatment with huAb A33. With subsequent antibody injections, detectable HAHA levels declined. HAHA of type II (17 % of patients) was characterized by a typically later onset of HAHA than in type I and by progressively increasing HAHA levels with each subsequent huAb A33 administration. The detectable onset of the immune response varied from patient to patient, ranging from week 2 to week 21 after the first administration of huAb A33. Colon cancer patients with type I HAHA did not develop infusion-related adverse events. In contrast, HAHA of type II was indicative of infusion-related adverse events necessitating termination of huAb A33 treatment. HAHA of type I was unexpected and has characteristics of a recall reaction, while HAHA type II has the characteristics of a newly induced immune response.

Many factors will influence the immunogenicity of a therapeutic antibody. Some are inherent to the antibody construct, while others are related to the way the antibody is administrated or the way the patient responds. To date, there is no adequate in vitro or in-silico method that can predict the in vivo immunogenicity of an antibody construct, and clinical trials are still the only way to assess whether an antibody is immunogenic or not. As HAHA responses in patients against an administered humanized antibody can come in different forms, it is critical to have a fast, specific and sensitive method for HAHA assessment that allows discrimination between response types which are harmful to a patient and require termination of treatment with antibody from those response types which appear tolerable and allow continued treatment with antibody. The BIACORE method used in this study fulfills these criteria and is now being utilized in all our studies with humanized antibodies.